Cycles in adversarial regularized learning
P Mertikopoulos, C Papadimitriou, G Piliouras - Proceedings of the twenty …, 2018 - SIAM
Regularized learning is a fundamental technique in online optimization, machine learning,
and many other fields of computer science. A natural question that arises in this context is …
and many other fields of computer science. A natural question that arises in this context is …
Optimistic mirror descent in saddle-point problems: Going the extra (gradient) mile
Owing to their connection with generative adversarial networks (GANs), saddle-point
problems have recently attracted considerable interest in machine learning and beyond. By …
problems have recently attracted considerable interest in machine learning and beyond. By …
Learning in games via reinforcement and regularization
P Mertikopoulos, WH Sandholm - Mathematics of Operations …, 2016 - pubsonline.informs.org
We investigate a class of reinforcement learning dynamics where players adjust their
strategies based on their actions’ cumulative payoffs over time—specifically, by playing mixed …
strategies based on their actions’ cumulative payoffs over time—specifically, by playing mixed …
Learning in games with continuous action sets and unknown payoff functions
P Mertikopoulos, Z Zhou - Mathematical Programming, 2019 - Springer
This paper examines the convergence of no-regret learning in games with continuous action
sets. For concreteness, we focus on learning via “dual averaging”, a widely used class of no…
sets. For concreteness, we focus on learning via “dual averaging”, a widely used class of no…
On the almost sure convergence of stochastic gradient descent in non-convex problems
In this paper, we analyze the trajectories of stochastic gradient descent (SGD) with the aim of
understanding their convergence properties in non-convex problems. We first show that the …
understanding their convergence properties in non-convex problems. We first show that the …
Bandit learning in concave N-person games
…, D Leslie, P Mertikopoulos - Advances in Neural …, 2018 - proceedings.neurips.cc
This paper examines the long-run behavior of learning with bandit feedback in non-cooperative
concave games. The bandit framework accounts for extremely low-information …
concave games. The bandit framework accounts for extremely low-information …
[PDF][PDF] No-regret learning and mixed Nash equilibria: They do not mix
…, T Lianeas, P Mertikopoulos… - arXiv preprint arXiv …, 2020 - proceedings.neurips.cc
Understanding the behavior of no-regret dynamics in general 𝑁-player games is a fundamental
question in online learning and game theory. A folk result in the field states that, in finite …
question in online learning and game theory. A folk result in the field states that, in finite …
Riemannian game dynamics
P Mertikopoulos, WH Sandholm - Journal of Economic Theory, 2018 - Elsevier
We study a class of evolutionary game dynamics defined by balancing a gain determined
by the game's payoffs against a cost of motion that captures the difficulty with which the …
by the game's payoffs against a cost of motion that captures the difficulty with which the …
On the convergence of gradient-like flows with noisy gradient input
P Mertikopoulos, M Staudigl - SIAM Journal on Optimization, 2018 - SIAM
In view of solving convex optimization problems with noisy gradient input, we analyze the
asymptotic behavior of gradient-like flows under stochastic disturbances. Specifically, we focus …
asymptotic behavior of gradient-like flows under stochastic disturbances. Specifically, we focus …
Learning with bandit feedback in potential games
…, J Cohen, P Mertikopoulos - Advances in Neural …, 2017 - proceedings.neurips.cc
This paper examines the equilibrium convergence properties of no-regret learning with
exponential weights in potential games. To establish convergence with minimal information …
exponential weights in potential games. To establish convergence with minimal information …